We use structured outputs a lot. But building those darn schemas to be ready for OpenAI is a real pain. So I built a simple service where I can post the prompt, or just my broken JSON, and it spits out a valid schema output for consumption the the LLM directly. Hope it helps other devs. Test it out, let me know if there's something I should improve. I'm here jamming on this today.